Mark error pages as such on Ctrl-Shift-T reload.

content/browser/frame_host/navigation_controller_impl.cc

Index: content/browser/frame_host/navigation_controller_impl.cc
diff --git a/content/browser/frame_host/navigation_controller_impl.cc b/content/browser/frame_host/navigation_controller_impl.cc
index 5c3f046e78f3aa954f1d220fb08fdcc5a0222ccb..4b21dee3d48fd1a11ceac826bd83084bf777800d 100644
--- a/content/browser/frame_host/navigation_controller_impl.cc
+++ b/content/browser/frame_host/navigation_controller_impl.cc
@@ -1126,6 +1126,8 @@ void NavigationControllerImpl::RendererDidNavigateToExistingPage(
   NavigationEntryImpl* entry = entries_[entry_index].get();
 
   // The URL may have changed due to redirects.
+  if (params.url_is_unreachable)
+    entry->set_page_type(PAGE_TYPE_ERROR);
   entry->SetURL(params.url);
   entry->SetReferrer(params.referrer);
   if (entry->update_virtual_url_with_url())
@@ -1178,6 +1180,8 @@ void NavigationControllerImpl::RendererDidNavigateToSamePage(
   existing_entry->set_unique_id(pending_entry_->GetUniqueID());
 
   // The URL may have changed due to redirects.
+  if (params.url_is_unreachable)
+    existing_entry->set_page_type(PAGE_TYPE_ERROR);
   if (existing_entry->update_virtual_url_with_url())
     UpdateVirtualURLToURL(existing_entry, params.url);
   existing_entry->SetURL(params.url);
@@ -1204,6 +1208,10 @@ void NavigationControllerImpl::RendererDidNavigateInPage(
   // entry and it will be the same page as the new navigation (minus the
   // reference fragments, of course).  We'll update the URL of the existing
   // entry without pruning the forward history.
+  // TODO(wjmaclean): do we need an else-clause here to clear the entry, since
+  // it is persisted between navigations?

[Charlie Reis, 2015/02/18 01:01:53]

This is a good question.  Can you test this manually?  I'd suggest navigating to
two pages, disabling your network, clearing cache, then going back.  That should
give you an existing entry with an error page.  Then repeat with the network
re-enabled.  I bet you're right that we would need to fix the page type to be
normal again (in all cases, not just this one).

[wjmaclean, 2015/02/18 19:10:11]

After a manual test, there doesn't seem to be a difference, though in a way I'm
not surprised given how few locations in the code test for PAGE_TYPE_ERROR. But,
I suspect it's the right thing to do, and so have added it in.

[Charlie Reis, 2015/02/18 23:53:04]

I think that's the right thing to do.  Thanks for being thorough!

+  if (params.url_is_unreachable)
+    existing_entry->set_page_type(PAGE_TYPE_ERROR);
   existing_entry->SetURL(params.url);
   if (existing_entry->update_virtual_url_with_url())
     UpdateVirtualURLToURL(existing_entry, params.url);